IPv6 Addressing ISP Workshops

Page created by Brent Fischer
 
CONTINUE READING
IPv6 Addressing ISP Workshops
IPv6 Addressing
                  ISP Workshops

These materials are licensed under the Creative Commons Attribution-NonCommercial 4.0 International license
(http://creativecommons.org/licenses/by-nc/4.0/)

                          Last updated 21st October 2017                                                      1
Acknowledgements
p   This material originated from the Cisco ISP/IXP Workshop
    Programme developed by Philip Smith & Barry Greene

p   Use of these materials is encouraged as long as the source is fully
    acknowledged and this notice remains in place

p   Bug fixes and improvements are welcomed
    n   Please email workshop (at) bgp4all.com

                                                    Philip Smith          2
Agenda
p   Recap: how it worked with IPv4
p   Getting IPv6 address space
p   Constructing a scalable IPv6 address plan
p   IPv6 addressing on LANs
p   IPv6 address plan example

                                                3
How it used to be
 Looking back at IPv4

                        4
How did it work for IPv4?
p   Up until 1994:
    n   Operators applied to InterNIC for address space
         p   1993 onwards: included RIPE NCC and APNIC
    n   Class A: Big organisations
    n   Class B: Medium organisations
         p   From 1992 onwards, multiple class Cs often handed out instead of single class Bs
    n   Class C: Small organisations
p   From 1994 onwards (classless Internet)
    n   Address space distributed by InterNIC (replaced by ARIN in 1998) and the
        other RIRs
    n   Distribution according to demonstrated need (not want)

                                                                                                5
IPv4 address plans (pre 1994)?
p   Prior to 1994, doing an address plan in IPv4 was very simple
p   Class C was used for one LAN
    n   If entity had more than one LAN, they’d normally get a class B
p   An organisation with a class B had 256 possible LANs
    n   And that was more than most networks had in those days
p   Organisations with more than 256 LANs tended to be Universities,
    big IT companies, etc
    n   They either had multiple class Bs, or even a class A

                                                                         6
Typical early 90s address plan
p   Organisation was not connected to the Internet as such
    n   But used TCP/IP internally
p   Would generally use 10.0.0.0/8
    n   Or any other class A that InterNIC had not handed out
p   10.X.Y.Z was a typical layout, where:
    n   X = building number
    n   Y = floor number
    n   Z = host address
    n   Where each subnet was a /24 (like a class C)
p   When these organisations connected to the Internet, they had to
    renumber
    n   Often into a /19 (the minimum allocation then)
                                                                      7
IPv4 address plans (post 1994)?
p   In the classful Internet days, IP address planning didn’t really exist
    n   The address space was big enough for most needs, as the number of devices
        and LANs were small
p   With the arrival of classless Internet, and IPv4 run out in the early
    90s
    n   IP address planning was needed
    n   Organisations got address space according to demonstrated need
         p   A previous class B might now only get a /19
         p   LANs no longer were automatically /24s
         p   etc

                                                                                    8
IPv4 address plans (post 1994)?
p   Advent of NAT assisted with delaying IPv4 runout
    n   End-user got single public address, and NATed on to that address
         p   (End-users could get lazy again)
p   Operators became more careful:
    n   RIR policy required “demonstrated need”
         p   Further allocations made only when existing allocations were proven to be mostly
             used up
    n   Started assigning address space across backbone according to the needs of
        the infrastucture
         p   No gaps, but still no real plan
         p   /30s for point-to-point links etc
         p   Although the “plans” often separated infrastructure address space from what went
             to customers
                                                                                                9
IPv4 address plans (today)
p   Chaotic?
p   Unstructured?
p   Undocumented?
p   With IPv4 address space almost all depleted
    n   Operators becoming ever more creative
    n   Operators extracting the last “drops” from their address space holdings
    n   It is a scramble just to keep network infrastructure addressed with public
        IPv4
    n   Some operators even use NAT within their backbones
    n   Some operators are reclaiming IPv4 address space loaned to their customers
    n   This hotch potch cannot and does not lead to good planning

                                                                                 10
IPv4 address plans (today)
p   More serious issues – because of the lack of structure, lack of
    planning:
    n   Infrastructure security filters become very hard to manage
         p   Adding yet another small block of IPv4 addresses to perimeter and control plane
             filters
    n   Traffic engineering is more challenging
         p   Lots of small blocks of address space to manage and manipulate
         p   With impacts on size of the global routing table too!
    n   Infrastructure addressing is difficult to manage
         p   Loopbacks and backbone point-to-point links no longer out of one contiguous
             block
    n   Access address pool resizing
         p   Broadband access pools renumbering, reassigning, etc
                                                                                               11
IPv6
p   IPv6 changes all this
p   Address space delegations are generous
    n   Reminders of the “old days” of classful IPv4
p   No NAT
p   Address planning is very possible
p   Address planning is very necessary
p   Documentation is very necessary
p   Operators accustomed to handling IPv4 in the 1980s and early
    1990s might be able to use those old skills for IPv6 !

                                                                   12
IPv6 Address Planning

                        13
Where to get IPv6 addresses
p   Your upstream ISP
p   Africa
    n   AfriNIC – http://www.afrinic.net
p   Asia and the Pacific
    n   APNIC – http://www.apnic.net
p   North America
    n   ARIN – http://www.arin.net
p   Latin America and the Caribbean
    n   LACNIC – http://www.lacnic.net
p   Europe and Middle East
    n   RIPE NCC – http://www.ripe.net/info/ncc
                                                  14
Internet Registry Regions

                            15
Getting IPv6 address space (1)
p   From your Regional Internet Registry
    n   Become a member of your Regional Internet Registry and get
        your own allocation
         p   Membership open to all organisations who are operating a network
    n   Address allocation policies listed on the individual RIR website
    n   Minimum allocation is a /32 (or larger if you will have more than
        65k /48 assignments)

                                                                                16
Getting IPv6 address space (2)
p   From your upstream ISP
    n   Receive a /48 from upstream ISP’s IPv6 address block
    n   Receive more than one /48 if you have more than 65k subnets
p   If you need to multihome:
    n   Apply for a /48 assignment from your RIR
    n   Multihoming with the provider’s /48 will be operationally
        challenging
         p   Provider policies, filters, etc

                                                                      17
Do NOT use 6to4
p   Some entities still use 6to4
    n   Deprecated in May 2015 due to serious operational and security issues
    n   Read https://tools.ietf.org/rfc/rfc7526.txt (BCP196) for the reasoning why

p   FYI: 6to4 operation:
    n   Take a single public IPv4 /32 address
    n   2002:::/48 becomes your IPv6 address block, giving 65k
        subnets
    n   Requires a 6to4 gateway
    n   6to4 is a means of connecting IPv6 islands across the IPv4 Internet

                                                                                     18
Nibble Boundaries
p   IPv6 offers network operators more flexibility with addressing
    plans
    n   Network addressing can now be done on nibble boundaries
         p   For ease of operation
    n   Rather than making maximum use of a very scarce resource
         p   With the resulting operational complexity
p   A nibble boundary means subnetting address space based on the
    address numbering
    n   Each number in IPv6 represents 4 bits = 1 nibble
    n   Which means that IPv6 addressing can be done on 4-bit boundaries

                                                                           19
Nibble Boundaries – example
p   Consider the address block 2001:db8:0:10::/61
    n   The range of addresses in this block are:
                      2001:0db8:0000:0010:0000:0000:0000:0000
                                        to
                      2001:0db8:0000:0017:ffff:ffff:ffff:ffff

    n   Note that this subnet only runs from 0010 to 0017.
    n   The adjacent block is 2001:db8:0:18::/61
                      2001:0db8:0000:0018:0000:0000:0000:0000
                                        to
                      2001:0db8:0000:001f:ffff:ffff:ffff:ffff

    n   The address blocks don’t use the entire nibble range

                                                                20
Nibble Boundaries – example
p   Now consider the address block 2001:db8:0:10::/60
    n   The range of addresses in this block are:
                        2001:0db8:0000:0010:0000:0000:0000:0000
                                          to
                        2001:0db8:0000:001f:ffff:ffff:ffff:ffff

    n   Note that this subnet uses the entire nibble range, 0 to f
    n   Which makes the numbering plan for IPv6 simpler
         p   This range can have a particular meaning within the ISP block (for example,
             infrastructure addressing for a particular PoP)

                                                                                           21
Addressing Plans – Infrastructure
p   All Network Operators should obtain a /32 from their RIR
p   Address block for router loop-back interfaces
    n   Number all loopbacks out of one /64
    n   /128 per loopback
p   Address block for infrastructure (backbone)
    n   /48 allows 65k subnets
    n   /48 per region (for the largest multi-national networks)
    n   /48 for whole backbone (for the majority of networks)
    n   Infrastructure/backbone usually does NOT require regional/geographical
        addressing
    n   Summarise between sites if it makes sense

                                                                                 22
Addressing Plans – Infrastructure
p   What about LANs?
    n   /64 per LAN
p   What about Point-to-Point links?
    n   Protocol design expectation is that /64 is used
    n   /127 now recommended/standardised
         p   http://www.rfc-editor.org/rfc/rfc6164.txt
         p   (reserve /64 for the link, but address it as a /127)
    n   Other options:
         p   /126s are being used (mimics IPv4 /30)
         p   /112s are being used
               § Leaves final 16 bits free for node IDs
         p   Some discussion about /80s, /96s and /120s too
         p   Some equipment doesn’t support /127s L
                                                                    23
Addressing Plans – Infrastructure
p   NOC:
    n   ISP NOC is “trusted” network and usually considered part of infrastructure
        /48
         p   Contains management and monitoring systems
         p   Hosts the network operations staff
         p   take the last /60 (allows enough subnets)
p   Critical Services:
    n   Network Operator’s critical services are part of the “trusted” network and
        should be considered part of the infrastructure /48
    n   For example, Anycast DNS, SMTP, POP3/IMAP, etc
         p   Take the second /64
         p   (some operators use the first /64 instead)

                                                                                     24
Addressing Plans – ISP to Customer
p   Option One:
    n   Use ipv6 unnumbered
    n   Which means no global unicast ipv6 address on the point-to-point link
    n   Router adopts the specified interface’s IPv6 address
         p   Router doesn’t actually need a global unicast IPv6 address to forward packets

         interface loopback 0
          ipv6 address 2001:db8::1/128
         interface serial 1/0
          ipv6 address unnumbered loopback 0

                                                                                             25
Addressing Plans – ISP to Customer
p   Option Two:
    n   Use the second /48 for point-to-point links
    n   Divide this /48 up between PoPs
    n   Example:
         p   For 10 PoPs, dividing into 16, gives /52 per PoP
         p   Each /52 gives 4096 point-to-point links
         p   Adjust to suit!
    n   Useful if ISP monitors point-to-point link state for customers
         p   Link addresses are untrusted, so do not want them in the first /48 used for the
             backbone &c
    n   Aggregate per router or per PoP and carry in iBGP (not ISIS/OSPF)

                                                                                               26
Addressing Plans – Customer
p   Customers get one /48
    n   Unless they have more than 65k subnets in which case they get a second /48 (and so
        on)
p   In typical deployments today:
    n   Several ISPs are giving small customers a /56 and single LAN end-sites a /64, e.g.:
         p   /64      if end-site will only ever be a LAN (not recommended!!)
         p   /56      for small end-sites (e.g. home/office/small business)
         p   /48      for large end-sites
p   Observations:
    n   This is a very active discussion area
    n   Don’t assume that a mobile endsite needs only a /64 – 3GPP Release 10 introduces
        DHCPv6-PD for tethering
    n   Some operators are distributing /60s to their smallest customers!!

                                                                                              27
Addressing Plans – Customer
p   Broadband Example:
    n   DHCPv6 pool is a /48
         p   DHCPv6 hands out /56 per customer
         p   Which allows for 256 customers per pool
    n   If BRAS has more than 256 customers, increase pool to a /47
         p   This allows for 512 customers at /56 per customer
    n   The whole nibble (/44) allows for 4096 delegations
    n   In all cases, BRAS announces entire pool as one block by iBGP

                                                                        28
Addressing Plans – Customer
p   Business “leased line”:
    n   /48 per customer
    n   One stop shop, no need for customer to revisit ISP for more addresses until
        all 65k subnets are used up
p   Hosted services:
    n   One physical server per vLAN
    n   One /64 per vLAN
    n   How many vLANs per PoP?
    n   /48 reserved for entire hosted servers across backbone
         p   Internal sites will be subnets and carried by iBGP

                                                                                      29
Addressing Plans – Customer
p   Geographical delegations to Customers:
    n   Network Operator subdivides /32 address block into geographical chunks
    n   E.g. into /36s
         p   Region 1: 2001:db8:1xxx::/36
         p   Region 2: 2001:db8:2xxx::/36
         p   Region 3: 2001:db8:3xxx::/36
         p   etc
    n   Which gives 4096 /48s per region
    n   For Operational and Administrative ease
    n   Benefits for traffic engineering if Network Operator multihomes in each region

                                                                                    30
Addressing Plans – Customer
p   Sequential delegations to Customers:
    n   After carving off address space for network infrastructure, Network Operator
        simply assigns address space sequentially
    n   Eg:
         p   Infrastructure:   2001:db8:0::/48
         p   Customer P2P:     2001:db8:1::/48
         p   Customer 1:       2001:db8:2::/48
         p   Customer 2:       2001:db8:3::/48
         p   etc
    n   Useful when there is no regional subdivision of network and no regional
        multihoming needs

                                                                                       31
Addressing Plans – Routing Considerations
p   Carry Broadband pools in iBGP across the backbone
    n   Not in OSPF/ISIS
p   Multiple Broadband pools on one BRAS should be aggregated if
    possible
    n   Reduce load on iBGP
p   Aggregating leased line customer address blocks per router or per
    PoP is undesirable:
    n   Interferes with ISP’s traffic engineering needs
    n   Interferes with ISP’s service quality and service guarantees

                                                                        32
Addressing Plans – Traffic Engineering
p   Smaller providers will be single homed
    n   The customer portion of the ISP’s IPv6 address block will usually
        be assigned sequentially
p   Larger providers will be multihomed
    n   Two, three or more external links from different providers
    n   Traffic engineering becomes important
    n   Sequential assignments of customer addresses will negatively
        impact load balancing

                                                                        33
Addressing Plans – Traffic Engineering
p   ISP Router loopbacks and backbone point-to-point links make up a
    small part of total address space
    n   And they don’t attract traffic, unlike customer address space
p   Links from ISP Aggregation edge to customer router needs one /64
    n   Small requirements compared with total address space
    n   Some ISPs use IPv6 unnumbered
p   Planning customer assignments is a very important part of
    multihoming
    n   Traffic engineering involves subdividing aggregate into pieces until load
        balancing works

                                                                                    34
Unplanned IP addressing
p   ISP fills up customer IP addressing from one end of the range:
                                        2001:db8::/32

                      12345

                ISP                   Customer Addresses
p   Customers generate traffic
    n   Dividing the range into two pieces will result in one /33 with all the customers and
        the ISP infrastructure the addresses, and one /33 with nothing
    n   No loadbalancing as all traffic will come in the first /33
    n   Means further subdivision of the first /33 = harder work

                                                                                           35
Planned IP addressing
p   If ISP fills up customer addressing from both ends of the range:
                                        2001:db8::/32

                    13579                     2 4 6 810

                 ISP Customer Addresses         Customer Addresses

p   Scheme then is:
    n   First customer from first /33, second customer from second /33, third from first
        /33, etc
p   This works also for residential versus commercial customers:
    n   Residential from first /33
    n   Commercial from second /33                                                         36
Planned IP Addressing
p   This works fine for multihoming between two upstream links (same
    or different providers)
p   Can also subdivide address space to suit more than two upstreams
    n   Follow a similar scheme for populating each portion of the address space
p   Consider regional (geographical) distribution of customer
    delegated address space
p   Don’t forget to always announce an aggregate out of each link

                                                                                   37
Addressing Plans – Advice
p   Customer address assignments should not be reserved or assigned
    on a per PoP basis
    n   Follow same principle as for IPv4
    n   Subnet aggregate to cater for multihoming needs
    n   Consider regional delegation
    n   ISP iBGP carries customer nets
    n   Aggregation within the iBGP not required and usually not desirable
    n   Aggregation in eBGP is very necessary
p   Backbone infrastructure assignments:
    n   Number out of a single /48
         p   Operational simplicity and security
    n   Aggregate to minimise size of the IGP
                                                                             38
Addressing Plans – Scheme
p   Looking at Infrastructure:
                                       2001:db8::/32

            /64             2001:db8:0::/48              /60   2001:db8:1::/48 to
                                                               2001:db8:ffff::/48

         Loopbacks       Backbone PtP & LANs          NOC       Customers

p   Alternative:
                                       2001:db8::/32

            /64      2001:db8:0::/48      /60 2001:db8:1::/48 2001:db8:2::/48 to
                                                               2001:db8:ffff::/48

         Loopbacks Backbone PtP          NOC     Customer       Customers           39
                      & LANs                        PtP
Addressing Plans
Planning
p   Registries will usually allocate the next block to be
    contiguous with the first allocation
    n   (RIRs use a sparse allocation strategy – industry goal is
        aggregation)
    n   Minimum allocation is /32
    n   Very likely that subsequent allocation will make this up to a /31
        or larger (/28)
    n   So plan accordingly

                                                                            40
Addressing Plans (contd)
p   Document infrastructure allocation
    n   Eases operation, debugging and management
p   Document customer allocation
    n   Customers get /48 each
    n   Prefix contained in iBGP
    n   Eases operation, debugging and management
    n   Submit network object to RIR Database

                                                    41
Addressing – Industry Best Practices
p   Industry has documented best practices for IPv6 address
    delegation from operators to end-users
p   It is what the majority of network operators are doing
    today:
    n   https://www.ripe.net/publications/docs/ripe-690/

p   Recommendation:
    n   Follow RIPE-690 guidelines – this will ensure long term
        scalability of your IPv6 deployment

                                                                  42
Addressing Tools
p   Examples of IP address planning tools:
    n   NetDot           netdot.uoregon.edu
    n   OpenNetAdmin     opennetadmin.com
    n   HaCi             sourceforge.net/projects/haci
    n   Racktables       racktables.org
    n   IPAT             nethead.de/index.php/ipat
    n   freeipdb         home.globalcrossing.net/~freeipdb/
p   Examples of IPv6 subnet calculators:
    n   ipv6gen          code.google.com/p/ipv6gen/
    n   sipcalc          www.routemeister.net/projects/sipcalc/

                                                                  43
IPv6 Addressing on LANs

                          44
IPv6 Addressing on LANs
p   Two options:
    n   Stateless Autoconfiguration (SLAAC)
         p   Client learns IPv6 address from the router on the subnet
    n   DHCPv6
         p   Client learns IPv6 address from a DHCP server (as for IPv4)

                                                                           45
SLAAC
p   IPv6 client learns address “from the LAN”
    n   Sends out “router solicit”
    n   Router responds with “router advertisement” containing subnet
        and default gateway
    n   Initial client state:
        Client:
        en3: flags=8863 mtu 1500
                    ether 68:5b:35:7d:3b:bd
                    inet6 fe80::6a5b:35ff:fe7d:3bbd%en3 prefixlen 64 scopeid 0x8

    n   Router does not need any specific configuration
        interface FastEthernet0/0
         ipv6 address 2001:db8:100::1/64
         ipv6 nd router-preference high
                                                                                   46
        !
SLAAC
p   On receiving response from the router:
    en3: flags=8863 mtu 1500
                ether 68:5b:35:7d:3b:bd
                inet6 fe80::6a5b:35ff:fe7d:3bbd%en3 prefixlen 64 scopeid 0x8
                inet6 2001:db8:100::6a5b:35ff:fe7d:3bbd prefixlen 64 autoconf
                inet6 2001:db8:100::18eb:2861:458e:862b prefixlen 64 autoconf temporary
                nd6 options=1

    Internet6:
    Destination        Gateway            Flags         Netif Expire
    default            fe80::219:30ff:fee UGc             en3

    n   Note the temporary address – this is the one used for all IPv6
        connectivity, and has a lifetime determined by the client’s
        operating system

                                                                                          47
DHCPv6
p   Works like DHCP on IPv4 infrastructure:
    n   DHCPv6 server distributes addresses from a pool on request
        from client
    n   DHCPv6 client configures IPv6 address
    n   Sample server configuration (Cisco IOS):
        ipv6 dhcp pool LABNET
         dns-server 2001:DB8:1::1
         dns-server 2001:DB8:2::2
         domain-name labnet
        !
        interface VLAN1
         ipv6 address 2001:DB8::1/64
         ipv6 dhcp server LABNET
        !
                                                                     48
Distributing subnets by DHCP
p   Two options:
    n   Static assignment (as in IPv4)
         p   Tell the customer what subnet they have
    n   DHCPv6-PD
         p   Use DHCPv6 Prefix Delegation feature to distribute subnets automatically

                                                                                        49
DHCPv6-PD
p   New for IPv6, is Prefix-Delegation (PD)
    n   Allows DHCP server to delegate subnets to clients
    n   Especially useful for Broadband deployments
    n   Server example on BRAS (Cisco IOS):
        ipv6 dhcp pool BB-CUST-1
         prefix-delegation pool BBCUST1 lifetime 1800 600
        !
        ipv6 local pool BBCUST1 2001:DB8:F00::/40 56
        !
        interface FastEthernet0/0
         ipv6 enable
         ipv6 dhcp server BB-CUST-1
        !

                                                            50
DHCPv6-PD
p   Client receives address delegation from the server:
    interface Dialer0
     description ADSL link to MY ISP
     ipv6 address autoconfig default
     ipv6 dhcp client pd ADSL-PD rapid-commit
    !
    interface Vlan1
     description Home Network
     ipv6 address ADSL-PD ::0:0:0:0:1/64
    !
    interface Vlan2
     description Home IP/TV Network
     ipv6 address ADSL-PD ::1:0:0:0:1/64
    !

    Vlan1 – IPv6 address: 2001:DB8:F00:3100::1/64
    Vlan2 – IPv6 address: 2001:DB8:F00:3101::1/64
                                                          51
Example Address Plan

                       52
Example Address Plan
p   Generic Network Operator
    n   Has 2001:db8::/32 address block
    n   Takes first /48 for network infrastructure
         p   First /64 for loopbacks
         p   Last /60 for NOC
    n   Takes second /48 for point to point links to customer sites
    n   Remainder of address space for delegation to customers,
        content hosting and broadband pools
p   Network Operator has 20 PoPs around the country

                                                                      53
Example: Loopback addresses
p   2001:db8:0::/48 is used for infrastructure
p   Out of this, 2001:db8:0:0::/64 is used for loopbacks
    n   Each loopback is numbered as a /128
p   Scheme adopted is:
    n   2001:db8::XXYY/128
         p   Where XX is the PoP number (01 through FF)
         p   Where YY is the router number (01 through FF)
    n   Scheme is good for:
         p   255 PoPs
         p   255 routers per PoP
         p   keeping addresses small/short
                                                             54
Loopbacks Example
      PoP 1  Loopbacks           PoP 10    Loopbacks
     Routers                     Routers
       cr1   2001:db8::101/128     cr1     2001:db8::a01/128
       cr2   2001:db8::102/128     cr2     2001:db8::a02/128
       br1   2001:db8::103/128     sr1     2001:db8::a05/128
       br2   2001:db8::104/128     sr2     2001:db8::a06/128
       sr1   2001:db8::105/128     ar1     2001:db8::a10/128
       sr2   2001:db8::106/128     ar2     2001:db8::a11/128
       ar1   2001:db8::110/128    gw1      2001:db8::a20/128
       ar2   2001:db8::111/128    gw2      2001:db8::a21/128
      gw1    2001:db8::120/128    etc…
      gw2    2001:db8::121/128
      etc…
                                                               55
Example: Backbone Point to Point links
p   Backbone Point to Point links come out of Infrastructure
    block 2001:db8:0::/48
    n   Scheme adopted is:
         p   2001:db8:0:MNXY::Z/64
    n   Where
         p   MN is the PoP number (01 through FF)
         p   XY is the LAN number (00 through 0F)
         p   XY is the P2P link number (10 through FF)
         p   Z is the interface address (0 or 1)
    n   Scheme is good for 16 LANs and 240 backbone PtP links per
        PoP, and for 255 PoPs
                                                                    56
LANs and PtP Links Example
    PoP 1      Subnet                       PoP 14     Subnet
    LAN1       2001:db8:0:101::/64           LAN1      2001:db8:0:e01::/64
    LAN2       2001:db8:0:102::/64           LAN2      2001:db8:0:e02::/64
    LAN3       2001:db8:0:103::/64           LAN3      2001:db8:0:e03::/64
    PtP1       2001:db8:0:111::/64           LAN4      2001:db8:0:e04::/64
    PtP2       2001:db8:0:112::/64           LAN5      2001:db8:0:e05::/64
    PtP3       2001:db8:0:113::/64           PtP1      2001:db8:0:e11::/64
    PtP4       2001:db8:0:114::/64           PtP2      2001:db8:0:e12::/64
    PtP5       2001:db8:0:115::/64           PtP3      2001:db8:0:e13::/64
    PtP6       2001:db8:0:116::/64           etc…
    PtP7       2001:db8:0:117::/64
    etc…

            Note: PtP links have /64 reserved but are addressed as /127s     57
Links to Customers
p   Some ISPs use “ip unnumbered” for IPv4 interface links
    n   So replicate this in IPv6 by using “ipv6 unnumbered” to address
        the links
    n   This will not require one /48 to be taken from the ISP’s /32
        allocation
p   Other ISPs use real routable addresses
    n   So set aside the second /48 for this purpose
    n   Gives 65536 possible customer links, assuming a /64 for each
        link

                                                                          58
Customer Links Example
     Customer                         Point to point link address
     Customer 1                            2001:db8:1:0::/64
     Customer 2                            2001:db8:1:1::/64
     Customer 3                            2001:db8:1:2::/64
     Customer 4 (link one)                 2001:db8:1:3::/64
     Customer 4 (link two)                 2001:db8:1:4::/64
     Customer 5                            2001:db8:1:5::/64
     Customer 6                            2001:db8:1:6::/64
     etc…

         Note1: PtP links are numbered out of 2001:db8:1::/48

         Note2: PtP links have /64 reserved but are addressed as /127s
                                                                         59
Example: Allocations from the /32
p   Master allocation documentation would look like this:
             Category     Purpose
             Single /64   Loopbacks
             Single /60   NOC
             Single /48   Backbone Point to Point links (/64 each)
             Single /48   Customer Point to Point links (/64 each)
             Single /40   65536 Broadband Customers in Region 1 (/56 each)
             Single /40   256 Enterprise Customers in Region 1 (/48 each)
             Single /40   65536 Broadband Customers in Region 2 (/56 each)
             Single /40   256 Enterprise Customers in Region 2 (/48 each)
             Etc…

                                                                             60
Example: Allocations from the /32
p   Detailed documentation:
               Address Blocks       Purpose

               2001:db8:0::/48      Infrastructure (Loops, NOC, PtP)

               2001:db8:1::/48      Customer Point to Point Links

               2001:db8:0110::/48   Customer One in Region 1

               2001:db8:0111::/48   Customer Two in Region 1

               2001:db8:0112::/48   Customer Three in Region 1

               2001:db8:1100::/40   Broadband Pool 1 in Region 1

               2001:db8:1200::/40   Broadband Pool 2 in Region 1

               2001:db8:8110::/48   Customer One in Region 2

               2001:db8:8111::/48   Customer Two in Region 2

               2001:db8:9100::/40   Broadband Pool 1 in Region 2

               2001:db8:9200::/40   Broadband Pool 2 in Region 2       61
Summary
p   First /48 for infrastructure
    n   Out of that, first /64 for Loopbacks
p   PoP structure within IPv6 addressing is very possible
    n   Greater flexibility than with IPv4
    n   Possible to come up with a simple memorable scheme
p   Documentation vitally important!

                                                             62
Presentation Recap
p   How it worked with IPv4
p   Getting IPv6 address space
p   Constructing a scalable IPv6 address plan
p   IPv6 addressing on LANs
p   IPv6 address plan example

                                                63
IPv6 Addressing
  ISP Workshops

                  64
You can also read